critical issues related to corrosion induced deterioration projection
TRANSCRIPT
Final Report
CRITICAL ISSUES RELATED TO CORROSION INDUCED DETERIORATION PROJECTION FOR CONCRETE MARINE BRIDGE SUBSTRUCTURES
submitted to
Florida Department of Transportation Research Center
605 Suwannee Street Tallahassee, Florida 32399
submitted by
William H. Hartt, Jingak Nam, and Lianfang Li Center for Marine Materials
Department of Ocean Engineering Florida Atlantic University – Sea Tech Campus
101 North Beach Road Dania Beach, Florida 33004
December 30, 2002
ii
EXECUTIVE SUMMARY
A series of 91 concrete G109 type specimens were exposed to cyclic wet-dry
ponding with either a 15 w/o NaCl solution or natural sea water for as long as four
years. Mix design variables included 1) cement alkalinity (equivalent alkalinities of
0.97 (high), 0.52 (normal and typical of cements used for Florida bridges), and 0.36
(low)), 2) water cement ratio (0.50, 0.41, and 0.37), 3) presence versus absence of
coarse aggregate, and 4) presence versus absence of fly ash. Corrosion potential
and macrocell current between bottom and top bars were periodically monitored in
order to determine the time at which active corrosion commenced. Subsequent to
corrosion initiation, specimens were autopsied and determinations made of 1) the
effective chloride diffusion coefficient, 2) pore water pH, and 3) the critical chloride
concentration for corrosion initiation was calculated.
The time for corrosion to initiate conformed to a distribution that was
represented by Weibull statistics. The mean time-to-corrosion for specimens
fabricated using the high alkalinity cement was approximately ten times greater than
for those fabricated from the normal and low alkalinity ones. Consequently, since
alkali silica reaction is generally not a problem in the State, specification of high
alkalinity cements for bridge construction should lead to significantly enhanced
resistance to corrosion induced concrete deterioration at no additional cost.
In cases where determinations were made, corrosion initiated at a macro-void
that happened to be present on the top side of the reinforcing steel. This is thought
to have occurred in conjunction with a corrosion cell established between the
contiguous concrete covered and void area steel, where the former served as a
cathode and the latter as an anode. On this basis, time-to-corrosion was influenced
by the size and density of such voids. The possibility is explored that additional
corrosion resistance could be realized by void control, as might be affected by
modified placeme nt procedures and inclusion of admixtures designed to reduce void
size and density.
iii
TABLE OF CONTENTS
Page
EXECUTIVE SUMMARY ………………………………………………………………………………………… ii
TABLE OF CONTENTS ………………………………………………………………………………………….. iii
INTRODUCTION .………………………………………………………………………………………………….. 1
Overview of Deterioration processes …………………………………………………………….. 1
General ………………………………………………………………………………………………………. 1 Corrosion Mechanism…………………………………………………………………………………. 1
Representation of Corrosion Induced Concrete Deterioration ………………………. 4
Representation of Corrosion Induced Concrete Deterioration ………………………. 6
General ……………………………………………………………………………………………………… 6 Pore Water pH ………………………………………………………………………………………….. 7 Chlorides ……………………………………………………………………………………………………. 8
PROJECT DESCRIPTION AND OBJECTIVES ……………………………………………………….. 10
EXPERIMENTAL PROCEDURE ……………………………………………………………………………… 10
Materials ……………………………………………………………………………………………………….. 10
Specimens …………………………………………………………………………………………………….. 10
Exposure and Specimen Analysis …………………………………………………………………. 11
RESULTS AND DISCUSSION ………………………………………………………………………………. 16
Time-to-Corrosion …………………………………………………………………………………………. 16
Weibull Analysis …………………………………………………………………………………………….. 20
Pore Water pH ………………………………………………………………………………………………… 27
Chloride Analysis Results ………………………………………………………………………………. 31
Diffusion Coefficients ……………………………………………………………………………….. 31 Chloride Threshold Concentrations …………………………………………………………. 33
ACKNOWLEDGEMENTS ………………………………………………………………………………………… 43
CONCLUSIONS …………………………………………………………………………………………………….. 43
RECOMMENDATIONS …………………………………………………………………………………………… 45
BIBLIOGRAPHY …………………………………………………………………………………………………….. 46
iv
LIST OF TABLES
Page
1. Compositional analysis results for each of the three cements ………………….. 11
2. Mix design for concrete specimens …………………………………………………………….. 12
3. Listing of specimens according to mix design ……………………………………………. 13
4. Specimen times-to-corrosion ………………………………………………………………………. 17
5. Listing of distribution parameters for the T i data ………………………………………. 24
6. Exposure time data for specimens that have become active ……………………. 35
7. Chloride analysis results for the voids identified in Figure 34 ………………….. 42
v
LIST OF FIGURES
Page
1. Schematic illustration of sea water migration and Cl- accumulation within a marine piping ………………………………………………………….. 4 2. Photograph of a cracked and spalled marine bridge piling …………………………. 5 3. Schematic illustration of the various steps in deterioration of reinforced concrete due to chloride induced corrosion …………………………… 5
1
INTRODUCTION
Overview of Concrete Deterioration Processes
General
While concrete has evolved to become the most widely used structural material
in the world, the fact that its capacity for plastic deformation is essentially nil
imposes major practical design limitations. This shortcoming is most commonly
overcome by incorporation of steel reinforcement into those locations in the concrete
where tensile stresses are anticipated. Consequently, concerns regarding
performance must not only focus upon properties of the concrete per se but also of
the embedded steel and, in addition, the manner in which these two components
interact. In this regard, steel and concrete are in most aspects mutually compatible,
as exemplified by the fact that the coefficient of thermal expansion for each is
approximately the same. Also, while boldly exposed steel corrodes actively in most
natural environments at a rate that requires use of extrinsic corrosion control
measures (for example, protective coatings for atmospheric exposures and cathodic
protection in submerged and buried situations), the relatively high pH of concrete
pore water (pH ≈ 13.0-13.8) promotes formation of a protective passive film such
that corrosion rate is negligible and decades of relatively low maintenance result.
Corrosion Mechanism
Disruption of the passive film upon embedded reinforcement and onset of active
corrosion can arise in conjunction with either of two causes: carbonation or chloride
intrusion (or a combination of these two factors). In the former case (carbonation),
atmospheric carbon dioxide (CO2) reacts with pore water alkali according to the
generalized reaction,
( ) OHCaCOCOOHCa 2322 +→+ , (1
which consumes reserve alkalinity and eventually reduces pore water pH to the 8-9
range, where steel is no longer passive. For dense, high quality concrete (for
2
example, high cement factor, low water-cement ratio, and presence of a pozzolanic
admixture) carbonation rates are typically on the order of one mm per decade or
less; and so loss of passivity from this cause within a normal design life is not
generally a concern. On the other hand, carbonation is often a problem for older
structures; first, because of age per se and, second, because earlier generation
concretes were typically of relatively poor quality (greater permeability) than more
modern ones.
Chlorides, on the other hand, arise in conjunction with deicing activities upon
northern roadways or from coastal exposure (or both). While this species (Cl-) has
only a small influence on pore water pH per se, concentrations as low as 0.6 kg/m3
(1.0 pcy) (concrete weight basis) have been projected to compromise steel passivity
(1,2). In actuality, it is probably not the concentration of chlorides per se that
governs loss of passivity but rather some function of the chloride-to-hydroxide ratio
([Cl-]/[OH-]), since the latter species (OH-) is prevalent in pore water and acts as an
inhibitor. This has been demonstrated by aqueous solution experiments from which
it is apparent that the Cl- threshold for loss of steel passivity increases with
increasing pH (3-8). On this basis, the relative amount of cement in a concrete mix
and cement alkalinity are likely to affect the onset of corrosion. Considerable
research effort has been focused upon identification of a chloride threshold;
however, a unique value for this parameter has remained illusive, presumably
because of the role of cement, concrete mix, environmental, electrochemical
potential, and reinforcement (composition and microstructure) variables that are
influential (9,10). Because Cl- and not carbonation induced loss of passivity is of
primary concern for modern bridge structures, subsequent focus in this report is
upon this corrosion cause.
Once steel in concrete becomes active, either in conjunction with chlorides
achieving a threshold concentration or pore solution pH reduction from carbonation
at the embedded steel depth, then the classical anodic iron reaction,
−+ +→ 2eFeFe 2 , (2
and cathodic oxygen reaction,
3
-2OH2eOHO2
122 →++ − , (3
occur at an accelerated rate. Despite the normally high alkalinity of concrete,
acidification may occur in the vicinity of anodic sites because of oxygen depletion and
hydrolysis of ferrous ions. Thus,
++ +→+ 2HFe(OH)O2HFe 222 . (4
The product H+ may be reduced and, along with O2 reduction at more remote
cathodic sites, further stimulate the anodic process. Irrespective of this, the net
reaction is
22
22 )Fe(OH2OHFeOHO2
1Fe →+→++ −+ (5
and, upon further oxidation,
3222 2Fe(OH)OHO2
12Fe(OH) →++ , (6
in addition to,
O3HOFe2Fe(OH) 2323 +→ , (7
as drying takes place.
Corrosion per se is seldom the cause of failure in reinforced concrete
components and structures. This arises because the final corrosion products (either
ferric oxide or hydroxide) have a specific volume that is several times greater than
that of the reactant steel; and accumulation of these in the concrete pore space
adjacent to anodic sites leads to development of tensile hoop stresses about the
steel which, in combination with the relatively low tensile strength of concrete
(typically 1-2 MPa), ultimately causes cracking and spalling.
4
Damage of this type has evolved to become a significant concern in the case of
coastal bridge sub-structures in Florida. Figure 1 illustrates the deterioration process
schematically where chlorides accumulate within the submerged zone from inward
sea water migration and in the atmospheric zone as a consequence of capillary flow,
splash, and spray. The lack of dissolved oxygen in the submerged zone precludes
Reaction 2, and so corrosion is rarely a problem here. However, ready availability of
both Cl- in the splash zone and of O2 at contiguous, more elevated locations results
in this location (splash zone) being particularly susceptible to corrosion induced
damage. Figure 2 shows a photograph of a marine bridge piling that exemplifies
this.
Figure 1: Schematic illustration of sea water migration and Cl- accumulation within a marine piping.
Representation of Corrosion Induced Concrete Deterioration
Corrosion induced deterioration of reinforced concrete can be modeled in terms
of three component steps: 1) time for corrosion initiation, 2) time, subsequent to
corrosion initiation, for corrosion propagation (appearance of a crack on the external
concrete surface), and 3) time for one or more surface cracks to develop into spalls
that require repair, rehabilitation, or replacement. Figure 3 illustrates these
schematically as a plot of cumulative damage versus time showing the above three
component times, T i, Tp, and Td (periods for initiation, propagation, continued
damage (spalling), respectively), as well as the time-to-failure, T f, or functional
service life (modified from Tutti (11)). Of the former three terms, T i typically
Concrete
Pile
Zone of Maximum Chlorides
Evaporative Water Loss
Inward and Upward Sea Water Flux
Sea Water
5
Figure 2: Photograph of a cracked and spalled marine bridge piling.
Figure 3: Schematic illustration of the various steps in deterioration of reinforced concrete due to chloride induced corrosion.
occupies the longest period; and so it is upon this parameter that corrosion control
measures generally focus. The approach adapted by the Florida Department of
Cum
ula
tive
Dam
age
Time
T i
Tf
Tp
Td
6
Transportation for new bridge construction is to extend the time -to-corrosion
initiation by a combination of 1) adequate concrete cover and 2) use of high
performance concretes; that is, concretes with permeability reducing (pozzolanic) or
corrosion inhibiting admixtures (or both). Likewise, the methods of Life-Cycle Cost
Analysis (LCCA) are employed to evaluate and compare different materials selection
and design options. This approach considers both initial cost and the projected life
history of maintenance, repair, and rehabilitation expenses that are required until
the design life is reached. These are then evaluated in terms of the time value of
money, from which Present Worth is determined. Comparisons between different
options can then be made on a normalized basis. Figure 4 graphically illustrates an
example life-cycle c ost scenario.
Figure 4: Schematic illustration of life cycle cost.
Analysis of Corrosion Initiation (Time-to-Corrosion)
General
Electrolyte composition is invariably a dominant factor affecting any corrosion
process, including that for steel in concrete. As noted above, the predominant
corrosion activator in concrete pore water is normally chlorides, whereas hydroxides
serve as a passivator. Consequently, T i for concrete structures is determined by the
Initial Cost
End of Functional Service Life
(Bridge Replacement)
Rehabilitation
Annual Maintenance
Repair
Cost
Time
7
competing influences of these two species. On the one hand, pore water pH is
affected by cement content, cement alkalinity, and exposures that promote
carbonation. On the other, chloride concentration, [Cl-], at the steel depth is
determined by 1) composition of this species in the environment, 2) its ingress rate
into concrete, and 3) concrete cover over the steel, with onset of active corrosion
defined by the Cl- threshold concentration for passive film breakdown, thc .
Pore Water pH
Corrosion experiments intended to simulate steel in concrete have historically
employed a saturated Ca(OH)2 solution, the pH of which is approximately 12.4.
However, with the advent of the pore water expression method (12-14) and
theoretical considerations, it was recognized that K+ and Na+ are the predominant
cations; and the solubility and concentration of these is such that a pH in excess of
13 typically occurs.
Limitations associated with pore water expression include, first, prior water
saturation of samples is required and, second, the method is more useful for pastes
and mortars since expression yields for concrete, particularly high performance ones,
is low. Consequently, both ex-situ and in-situ leaching methods (15,16) have also
been developed, where the former involves exposure of a powder sample to distilled
water and the latter placement of a small quantity of water into a drilled cavity in
hardened concrete. A limitation in the case of ex-situ leaching is that solid Ca(OH)2
from the concrete becomes dissolved and elevates [OH-] compared to what
otherwise would occur. Also, the dissolved Ca(OH)2, if saturated, buffers the
leachate at a pH of about 12.4. These limitations are minimized by the in-situ
method because only about 0.4 mL of distilled water is employed; however, water
saturation of the specimen is required here also. Recently, a modification of the ex-
situ method was proposed whereby a correction is made for the [OH-] resulting from
Ca(OH)2 dissolution (7,17); however, solubles in unreacted cement partic les may
also become dissolved, thereby elevating the calculated pH compared to what
actually existed in the pore water. Consequently, this procedure may be more a
measure of inherent alkalinity than of pore water pH.
8
Chlorides
The mechanism of Cl- intrusion into concrete invariably involves both capillary
suction and diffusion; however, for situations where the depth to which the former
(capillary suction) occurs is relatively shallow compared to the reinforcement cover,
diffusion alone is normally considered. Analysis of the latter (diffusion) is
accomplished in terms of Fick’s second law or,
( ) ( )
∂
∂⋅
∂
∂=
∂
∂
x
Tx,cD
tt
Tx,c, (8
where c(x,T) is [Cl-] at depth x beneath the exposed surface after exposure time T
and D is the diffusion coefficient. As Equation 8 is written, D is assumed to be
independent of concentration. The solution in the one-dimensional case is,
⋅
−=−
−
TD2
xerf1
cc
cT)c(x,
os
o , (9
where
co is the initial or background Cl- concentration in the concrete,
cs is the Cl- concentration on the exposed surface, and
erf is the Gaussian error function.
Assumptions involved in arriving at this solution are, first, cs and D are constant with
time and, second, the diffusion is “Fickian;” that is, there are no Cl- sources or sinks
in the concrete. In actuality, cs may increase with exposure time, although steady-
state values in the range 0.3-0.7 (percent of concrete weight) have been projected
to result after about six months (18). Factors that effect cs have been projected to
include type of exposure, mix design (cement content, in partic ular), and curing
conditions (19). Also, the diffusion coefficient that is calculated from Equation 9 is
an “effective” value, Deff, since it is weighted over the relevant exposure period due,
first, to the fact that cs may vary and, second, because of progressive cement
hydration with time, and, lastly, chemical and physical Cl- binding such that the
concrete acts as a sink for this species and some fraction is no longer able to diffuse.
9
In the approach represented by Equation 9, c(x,T), co, and cs are measured
experimentally (normally by wet chemistry analysis) and Deff is calculated based
upon knowledge of reinforcement cover and exposure time. Experimental scatter
and error may be minimized by measuring c(x,T) at multiple depths and employing a
curve-fitting algorithm to calculate Deff. Also, if Deff is known from one sampling set,
then thc can be determined by measuring c(x,T) at the reinforcement depth (c rd) at
the time and location of corrosion initiation and solving Equation 9 recognizing that
for this situation, c rd ≈ c th. In any case, the parameters that affect Cl- intrusion rate
are cs, and Deff, both of which depend upon exposure conditions such as relative
humidity and time-of-wetness and material variables such as composition and
microstructure.
A factor that is not normally considered in time-to-corrosion evaluations is that
cs, x, c th, and Deff are statistically distributed parameters. Consequently, T i also
conforms to some distribution. This is important because decisions regarding repairs
and rehabilitations are often made based upon damage having occurred over some
percentage of the structure. Alternatively, loss of a single critical, non-redundant
structural element (substructure bridge pier, for example) causes failure.
The approach of employing pozzolanic and corrosion inhibiting admixtures that
has been adapted by the FDOT for enhancing concrete durability, as noted above,
can be related to and interpreted in terms of Equation 9. In this regard, the most
prevalent corrosion inhibitor (Ca(NO2)2), in effect, increases cth, whereas pozzolans
(fly ash, silica fume, and blast furnace slag) reduce Deff. Surface [Cl-] may also be
effected by pozzolans.
It is generally recognized that cth increases with increasing cement alkalinity
(10), although no systematic investigation that addresses this has been performed.
For some geographical locations, concerns regarding alkali-silica reaction (ASR),
whereby aggregates swell from a chemical reaction with alkaline pore water,
preclude utilizing high alkalinity cements, since ASR should occur at a greater rate
the higher the pore water alkalinity. The native Florida limestones that serve as
coarse aggregate for concrete bridge construction in most of the State are not
susceptible to ASR, however. Consequently, if it can be demonstrated that c th for
high alkalinity cements exceeds that for normal ones, then use of these cements
10
should yield concrete structures with enhanced resistance to corrosion induced
deterioration but with no additional cost or construction complexity.
PROJECT DESCRIPTION AND OBJECTIVES
The objectives of this project were to evaluate the extent to which high alkalinity
cement in concrete elevates the chloride concentration threshold for breakdown of
the passive film upon steel reinforcement and active corrosion initiation and, based
upon this, determine the benefits that might be realized in terms of extending time-
to-corrosion by using such cements in Florida coastal bridges. The effort involved
fabrication of 91 G109 specimens (20) for which mix design variables included 1)
cements of three alkalinities, 2) three water-to-cement ratios, 3) presence versus
absence of fly ash, and 4) presence versus absence of coarse aggregate. An
additional variable was that some specimens were exposed to 15 w/o NaCl and
others to natural sea water.
EXPERIMENTAL PROCEDURE
Materials
As noted above, cements of three alkalinities were employed in fabricating the
concrete specimens. Table 1 lists the composition for each of these cements. The
normal alkalinity one is typical of what is currently employed in bridge construction
in Florida. The equivalent alkalinity (EqA) for the three cements was determined
from the equation,
OKw/o0.658Ow/oNaEqA 22 ⋅⋅+= . (10
This revealed values of 0.355 for the “low” alkalinity cement (LA), 0.519 for the
“normal” (NA), and 0.972 for the “high” (HA). Table 2 lists the different mix designs.
Specimens
Seven specimens were fabricated for each mix design based upon selected combinations of
cement type, water-cement ratio (w/c), and presence versus absence of fly ash and coarse
aggregate, where four of these specimens contained reinforcement according to the
11
Table 1: Compositional analysis results for each of the three cements. VALUE, percent COMPOUND LOW NORMAL HIGH SiO2 21.93 21.88 20.63 Al2O3 5.16 5.64 4.44 Fe2O3 3.70 3.87 2.56 CaO 65.02 64.42 63.39 MgO 1.39 0.98 3.85 SO3 2.38 2.87 3.96 Na2O 0.099 0.164 0.192 K2O 0.39 0.54 1.19 TiO2 0.29 0.31 0.22 P2O5 0.209 0.123 0.063 Mn2O3 0.049 0.038 0.058 SrO 0.09 0.064 0.041 Cr2O3 0.003 0.002 0.003 C3A 7.43 8.39 7.43 C2S 24.2 29.3 16.5 C3S 51.255 44.414 56.461 Equivalent Alkalinity 0.355 0.519 0.972
standard G109 design (20) and three were blanks (no reinforcement). Diameter of
the reinforcing steel bars was 12.7 mm (#4), and surface preparation involved wire
brushing just prior to placement. Table 3 provides a listing of specimens according
to the different test categories and indicates individual specimen numbers. The top
bar of each reinforced specimen was electrically connected to the two bottom bars
through a 100-Ohm shunt and a switch that allowed the circuit to be opened or
closed. Figure 5 shows the specimen configuration schematically.
Exposure and Specimen Analyses
A one week wet-one week dry top surface ponding cycle began on June 3, 1998
using a 15 w/o sodium chloride solution for all specimens except for numbers 85-91
for which natural sea water was employed. At the end of the drying cycle, potential
and macro-cell current was measured both from the voltage drop across the shunt
and by a zero resistance ammeter with the shunt by-passed. Active corrosion was
defined initially as having commenced if, for two consecutive data acquisition
periods, the macro-cell current was 10 µA or greater and potential for the top bar
was –0.28 VSCE or more negative. Later, this definition was relaxed; and active
12
Table 2: Mix design for concrete specimens.
Specimen Designation HA,NA,LA-0.37 HA,NA,LA-0.41 HA,NA,LA-0.50 NAF-0.37 NAF-0.41 Batch Materials kg/m3 (lb/y3) kg/m3 (lb/y3) kg/m3 (lb/y3) kg/m3 (lb/y3) kg/m3 (lb/y3)
Cement (Type II) 388.4 (657.4) 388.4 (657.4) 388.4 (657.4) 310.7 (526.6) 310.7 (526.6) Water 143.5 (243.2) 159.0 (269.5) 193.9 (328.7) 143.5 (243.2) 159.0 (269.5)
Fine Aggregate 742.9 (1259.1) 702.0 (1189.8) 610.1 (1034.0) 742.9 (1259.1) 702.0 (1189.8) Coarse Aggregate 980.1 (1661.1) 980.1 (1661.1) 980.1 (1661.1) 980.1 (1661.1) 980.1 (1661.1) Fly Ash (Class F) 0 0 0 77.6 (131.5)* 77.6 (131.5)*
Admixtures kg/cwt (oz/cwt) kg/cwt (oz/cwt) kg/cwt (oz/cwt) kg/cwt (oz/cwt) kg/cwt (oz/cwt)
WRDA-64 0.51 (6.00) 0.51 (6.00) 0.51 (6.00) 0.51 (6.00) 0.51 (6.00) WRDA-19 1.70 (20.00) 1.28 (15.00) 0 1.70 (20.00) 1.28 (15.00)
Batch Target Properties** Slump, mm (in) 76 (3) 76 (3) 200 (3) 76 (3) 77 (3) Unit Weight, kg/m3 (lb/ft3) 2,254.1 (141.5) 2,228.6 (139.9) 2,171.3 (136.3) 2,254.1 (141.5) 2,228.6 (139.9) Water-Cement Ratio 0.37 0.41 0.50 0.37 0.41
* Twenty percent cement replacement. ** Air content for all mixes was 4 percent.
13
Table 3: Listing of specimens according to mix design. Specimen No. Reinforcement Cement Alk. w/c 01-04, 85-88* Yes 05-07, 89-91* No
High 0.37
08-11 Yes 12-14 No
High 0.41
15-18 Yes 19-21 No
High 0.5
22-25 Yes 26-28 No
Normal 0.37
29-32 Yes 33-35 No
Normal 0.41
36-39 Yes 40-42 No
Normal 0.5
43-46 Yes 47-49 No
Low 0.37
50-53 Yes 54-56 No
Low 0.41
57-60 Yes 61-63 No
Low 0.5
64-67 Yes 68-70 No
HAG** 0.50
71-74 Yes 75-77 No
NAF*** 0.37
78-81 Yes 82-84 No
NAF*** 0.41
* Ponded with natural sea water. All other specimens ponded with 15% NaCl.
** High alkalinity mortar (no coarse aggregate). *** Normal alkalinity cement with 20% fly ash replacement.
corrosion was defined solely in terms of potential having decreased to or become
more negative than -0.28 VSCE. After corrosion initiation, specimens were autopsied
and the corrosion state assessed. This involved making a saw-cutting lengthwise to
a depth of about 50 mm along both sides opposite the top bar and then splitting
open the specimen. At certain intervals, a core was taken from within the ponded
area of blank specimens (no reinforcing steel) of the different mixes. The cores
were then sliced and the individual slices ground to powder. In addition, powder
drillings were acquired along the top side of the upper rebar trace at both the
corrosion site and elsewhere where the steel remained passive.
Chloride analyses were performed using the FDOT acid soluble method (21).
14
Figure 5: Geometry of the G109 type specimens.
The profile data acquired from the blank specimens were employed in a least-
squares curve-fitting algorithm to Equation 9, from which Deff was calculated.
Measurement of pore water pH for samples of each cement alkalinity and w/c
POND 25
64
114
25
25
152
CONCRETE SPECIMEN
13
13
CONCRETE SPECIMEN
POND
152 203 280 381
25
25
152
TAPE REBAR
REBAR
15% NaCL SOLUTION
13
13
All Dimensions
in mm
15
was performed using an ex-situ leaching procedure that was developed in
conjunction with FDOT Project WPI 0510716 (7,17). This involved retrieving 5 cm
diameter by 15 cm long cores from the Cl--free portion of representative specimens.
The outer one cm thick layer was discarded to avoid concrete that was possibly
carbonated, and the core was then pulverized until all material passed a #50 sieve.
The concrete powder was divided into approximately 50 g increments with each
being stored in an individual air-tight plastic bottle.
For leaching, 50 ml of de-ionized water was added to the individual samples; and
then these were mixed and allowed to sit for three days. Separate experiments
determined that this leaching time was sufficient for equilibrium to be approached
(22). The samples were then filtered to yield about 30 g of liquid. This was divided
into two approximately equal parts, each of which was titrated.
The initial titration was for OH- concentration using 0.1N HCl. This involved
measurement of the potential of a combination pH-electrode in the titrated solution
and determination of the end-point noted. Immediately after the OH- titration, pH of
the same solution was first adjusted and then titrated for Ca2+ concentration using
0.01N EDTA and hydroxy naphthol blue as an indicator. The end point was defined
by a change of the solution color from purple to blue. The second solution was
titrated identically to the first, and results for the two pairs of titrations were then
averaged.
The determination of pore water pH was based upon the following assumptions:
1. The total amount of alkali ions measured in the leachant equals that in the
concrete pore water. This assumes that alkali ions have no binding effect
similar to that of chlorides.
2. The concrete is fully hydrated.
3. The OH- in the leachant is balanced by the Na+ and K+.
On the basis of the third assumption,
16
l2
lp ][Ca2][OH][OH +−− ⋅−= , (11
where [OH-]p is the concentration of this ion in the leachant that was associated with
K+ and Na+ in the pore water; and [OH-]l and [Ca2+]l are the respective
concentrations in the leachant, as determined from the titrations. Thus, the [OH-] in
concrete pore water ([OH-]pw ) can be calculated as,
?d
WV][OH
][OH
concr
concr
OHlpw
2
⋅
⋅=
−− , (12
where
Wconcr is the total weight of concrete powder used for leaching (g),
VH2O is the total amo unt of water used for leaching (ml),
? is concrete porosity (vol. fraction), and
dconcr is concrete density, (assumed as 2.3g/cm3).
The concrete porosity was determined according to ASTM C642-90 (23) for concrete
water absorption capacity.
Finally, the concrete pore water pH was calculated from the expression,
pH=14+log (?[OH-]pw), (13
where ? is the activity coefficient for OH-, which is approximately 0.7 when [OH-]
exceeds 0.1 mol/L.
RESULTS AND DISCUSSION Time-To-Corrosion
At the conclusion of the experiments (July, 2002), 28 of the 52 reinforced G109
specimens had exhibited active corrosion. Table 4 provides a listing of all specimens
and of the T i values in the cases of corrosion activity. In instances where Ti was
relatively brief, the shift in corrosion potential, φcorr, from the range indicative of
passivity (typically φcorr more positive than approximately –0.10 VSCE) to that of
17
Table 4: Specimen times-to-corrosion.
Specimen Mix Time to Corr. Specimen Mix Time to Corr. Number Design (days) Number Design (days)
1 HA 0.37 >1,600 43 LA 0.37 >1,600 2 HA 0.37 >1,600 44 LA 0.37 >1,600 3 HA 0.37 >1,600 45 LA 0.37 190 4 HA 0.37 >1,600 46 LA 0.37 >1,600 8 HA 0.41 >1,600 50 LA 0.41 297 9 HA 0.41 1,329 51 LA 0.41 >1,600 10 HA 0.41 1,192 52 LA 0.41 567 11 HA 0.41 1,401 53 LA 0.41 >1,600 15 HA 0.50 136 57 LA 0.50 107 16 HA 0.50 297 58 LA 0.50 234 17 HA 0.50 391 59 LA 0.50 136 18 HA 0.50 225 60 LA 0.50 136 22 NA 0.37 >1,600 64 HAG 0.50 149 23 NA 0.37 >1,600 65 HAG 0.50 178 24 NA 0.37 1,096 66 HAG 0.50 107 25 NA 0.37 1,192 67 HAG 0.50 163 29 NA 0.41 136 71 NAF 0.37 >1,600 30 NA 0.41 206 72 NAF 0.37 >1,600 31 NA 0.41 339 73 NAF 0.37 >1,600 32 NA 0.41 1,120 74 NAF 0.37 >1,600 36 NA 0.50 93 78 NAF 0.41 >1,600 37 NA 0.50 93 79 NAF 0.41 >1,600 38 NA 0.50 136 80 NAF 0.41 >1,600 39 NA 0.50 178 81 NAF 0.41 >1,600
active corrosion (φcorr more negative than –0.28 VSCE) (24) was relatively abrupt.
Concurrently, macro-cell current increased from negligible to greater than 10 µA. In
the case of specimens for which corrosion initiated after a relatively long exposure
time, this change was more gradual with specimens often reverting to the passive
state. Figure 6 presents potential versus time data for specimens of the LA0.50 mix
design, which exemplifies the former type of behavior. Figure 7 presents
comparable data for Specimen 24, which typifies the latter. None of the fly ash or
sea water ponded specimens have become active. In the former case, this
presumably resulted because of reduced Deff and in the latter because of lower [Cl-]
in the sea water compared to 15 w/o NaCl.
18
-600
-500
-400
-300
-200
-100
0
0 50 100 150 200 250 300
Exposure Time, days
Pote
ntial
, V (
SCE)
57
58
59
60
(a)
0
50
100
150
200
0 50 100 150 200 250 300
Exposure Time, days
Macr
o-C
ell C
urr
ent,
mA
57
58
59
60
(b)
Figure 6. Potential (a) and macro-cell current data (b) for specimens of mix
design LA0.50.
19
-400
-300
-200
-100
0
0 200 400 600 800 1000 1200
Exposure Time, days
Pote
ntial
, V (
SCE)
(a)
0
2
4
6
8
10
0 200 400 600 800 1000 1200
Exposure Time, days
Mac
ro-C
ell Curr
rent,
mA
(b)
Figure 7: Potential (a) and macro-cell current (b) versus exposure time for
Specimen Number 24.
20
Figure 8 shows the appearance of corrosion on Specimen Number 59 (LA0.50)
which became active relatively early (T i = 128 days). Termination of exposure and
sectioning occurred approximately 20 days subsequent to disclosure of corrosion
activity. In this case, the attack was relatively advanced, presumably because of 1)
comparatively low concrete resistivity, as affected by the 0.50 w/c, and 2) the
corrosion had been ongoing for several weeks. Correspondingly, Figure 9 shows a
photograph of corrosion on Specimen Number 11 (HA0.41) and reveals that the
extent of attack in this case was modest by comparison. Note that this latter
specimen experienced a potential and macrocell current reversion subsequent to
initially becoming active. For specimens in the second category, corrosion was
considered to have initiated at the time when potential first dropped below –0.28
VSCE, irrespective of any subsequent, more positive potential drift. Macro-cell current
was often less that 10 µA at this time.
Weibull Analysis
Most commonly, experimental corrosion programs focus upon mean time-to-
corrosion for a specific set of test variables. However, T i is statistically distributed;
and data thereof should be treated accordingly. For example, a bridge subjected to
deicing salts may become functionally obsolete when 20 percent of the deck is
delaminated or spalled. Relatedly, corrosion initiates upon substructure piers,
followed by concrete cracking and spalling, according to some distribution. Unlike
the bridge deck case, however, pier deterioration is more critical since achievement
of a single limit state constitutes closure or failure of the entire structure. Of
particular concern in the latter case are situations where the distribution has a
relatively large standard deviation. Here, the average or typical degree of distress
may be modest; but the deterioration state of the initially corroded one may be
advanced.
Weibull anslysis is a commonly employed engineering and scientific tool for
representing and projecting failures, even when relatively small samples are
involved. The two parameter Weibull distribution is characterized by the expression,
ß
?T
e1F(T)
−
−= , (14
21
Figure 8: Photograph of section halves of Specimen Number 59 (LA0.50) showing severe localized corrosion near the center of the rebar length.
Figure 9: Sectioned view of Specimen 11-HA 0.41 showing corrosion on the upper half of the top (anodic) bar and corrosion products on the bar trace.
Saw Cut
Rebar
Rebar Trace Concrete Fracture
Corrosion
22
where
F(T) = the cumulative distribution function (CDF) of the failed (corroded) fraction,
T = failure time (T i in the present analysis), β = the slope or shape parameter, and η = characteristic life or scale parameter (time at which 63.2 percent of the
population fails (corrodes).
The mean time to failure, Tm, is related to η by the expression,
+⋅=
ß
11G?Tm , (15
where ( )•Γ is the gamma function. On the other hand, the value for β can infer
mechanistic information. Thus, β < 1 indicates “infant mortality” or early-age
failures as can occur as a consequence of manufacturing defects. An example in this
category could be concrete which exhibits cracks at the time of construction such
that chlorides have rapid, early access to the reinforcement. Situations where β = 1
constitute random (time independent) failures, whereas failures with β > 1 are
indicative of wear-out. A critical, non-redundant component, the failure of which is
characterized by a small β, is of particular concern. As an example, the initial pier to
fail on a bridge with 100 piers, 50 years mean time-to-failure, and β = 1 is projected
to occur after approximately eight months. However, if β = 4, the initial failure
occurs at year 17.
An additional advantageous feature of Weibull analysis is that it incorporates
run-outs, termed “suspensions” (specimens for which corrosion has not yet initiated
in the present case).
The data in Table 4 for specimen groups that exhibited at least one incident of
active corrosion were represented using the two-parameter Weibull analysis. Thus,
Figures 10-12 show CDF plots for mix designs of each of the three cement alkalinities
and w/c = 0.50, 0.41, and 0.37, respectively. Values for η, β, the coefficient of
determination (r2), the total number of specimens in each group (n, four in all
23
Figure 10: Two-parameter Weibull distribution of the T i data for specimens of the 0.50 mix design.
Figure 11: Two-parameter Weibull distribution of the T i data for specimens of the 0.41 mix design. Beta for the HA data (1.40) was forced as average for the other two cement alkalinities.
η β r2 n/s 292 3.16 0.98 4/0 171 4.34 0.86 4/0 200 4.16 0.77 4/0 196 5.84 0.95 4/0
η β r2 n/s 512 1.31 0.86 4/0 934 1.59 1.00 4/2 2598 1.40 4/1
24
Figure 12: Two-parameter Weibull distribution of the T i data for specimens of the 0.37 mix design. Beta values were forced as 4.00.
cases), and the number of suspensions (s) for each mix design are also indicated.
Values for these are tabulated collectively in Table 5. Clearly, the lack of specimens
for which corrosion has initiated in the lowest w/c case precludes interpretation here
other than to reaffirm the beneficial effect of low w/c. In the 0.41 w/c case, β was
forced as unity (the closest integer β value for NA specimens) for the LA case, where
Table 5: Listing of distribution parameters for the T i data. Mix Design w/c Alkalinity
η β r2 n/s
HAG 196 5.84 0.95 4/0 HA 292 3.16 0.98 4/0 NA 171 4.34 0.86 4/0
0.50
LA 200 4.16 0.77 4/0 HA 2,598 1.40* -‘ 4/1 NA 511 1.31 0.86 4/0
0.41 LA 934 1.59 1.00** 4/2
HA 3,205 4.00** - 4/4 NA 1,541 4.00** - 4/2
0.37 LA 1,959 4.00** - 4/3
* Assigned average for other two cement alkalinities. ** Beta value forced.
η β r2 n/s 1541 4.00 1.00 4/2 1959 4.00 1.00 4/3 3205 4.00 1.00 4/4
25
only one specimen has become active. Likewise, Figures 13-15 present the same
data as in Figures 10-12 but according to cement alkalinity rather than w/c. These
Figure 13: Two-parameter Weibull distribution of the T i data for specimens of the HA cement. Beta values were forced as 4.00.
Figure 14: Two-parameter Weibull distribution of the T i data for specimens of the NA cement. Beta values were forced as 4.00.
η β r2 n/s 283 4.00 0.98 4/0 1759 4.00 0.85 4/1 2306 4.00 1.00 4/4
η β r2 n/s 173 4.00 0.86 4/0 401 4.00 0.86 4/0 1541 4.00 1.00 4/2
26
Figure 15: Two-parameter Weibull distribution of the T i data for specimens of the LA cement. Beta values were forced as 4.00.
results indicate that, first, Ti increased with decreasing w/c, as has been documented
historically by numerous investigators and, second, T i was highest for the HA
specimens. Also, η for the HA specimens with coarse aggregate was about 32
percent greater than for the HAG specimens (no course aggregate) (Figure 10). This
is consistent with the inhibiting effect of coarse aggregate upon Cl- intrusion, as
reported in a companion study (25). A generalized trend of increasing T i with
increasing ceme nt alkalinity was not disclosed, however, since for each w/c, η was
greater for the LA specimens than for the NA.
Eta values, including projections where no specimens have yet become active,
are shown as a function of cement alkalinity in Figure 16. Again, caution must be
exercised with regard to interpretation because of the limited number of specimens
that have become active; however, based upon these results, Ti for the HA
specimens exceeded that for the NA and LA by a factor or 1.7-5.1.
The finding that η for the LA mix design exceeded that of the NA ones may
relate to reduced Cl- binding with increasing OH- over the pore water pH range that
applies here. In this regard, it has been reported that, because of this effect, a
η β r2 n/s 201 4.00 0.77 4/0 585 4.00 1.00 4/2 1958 4.00 1.00 4/3
27
0.00
0.25
0.50
0.75
1.00
1.E+02 1.E+03 1.E+04Eta, days
Cem
ent
Alk
alin
ity,
per
cent
w/c = 0.50w/c = 0.41w/c = 0.37
Figure 16: Plot of the eta versus cement alkalinity data.
decrease in pH at a fixed Cl- concentration can result in an decrease in Cl--to-OH-
ratio (26,27).
Pore Water pH
Results of the concrete porosity determinations based on water absorption
capacity are shown in Figure 17. These data indicate that porosity increased with
increasing w/c but was not significantly affected by cement alkalinity. The two
concretes containing fly ash (NAF0.37 and NAF0.41) did not yield a lower porosity
than the comparable non-admixed concretes (NA0.37 and NA0.41). This may be due
to insufficient concrete curing in a highly humid environment.
Figure 18 shows the measured pore water [OH-] for the various concrete mixes
and indicates that this parameter was not sensitive to w/c. However, pore water
[OH-] increased with increasing cement alkalinity. Figure 19 plots these same results
in terms of cement equivalent alkalinity. The data indicate that the average pore
water [OH-] was approximately 0.17 M for the LA concrete, 0.25 M for the NA, and
0.55M for HA.
28
Figure 17: Plot of concrete porosity versus w/c for specimens of the different mix designs.
Figure 18: Pore water [OH-] for the different concrete mixes as a function of w/c.
0.35 0.40 0.45 0.50 0.55
Water-to-Cement Ratio
0.10
0.15
0.20
Poro
sity
, vo
l. f
ract
ion
LA
NA
HA
NAF
HAG
0.0
0.10
0.20
0.30
0.40
0.50
0.60
0.70
0.35 0.40 0.45 0.50 0.55
Water-to-Cement Ratio
Pore
Wate
r [O
H- ]
,M
LA
NA
HA
NAF
HAG
29
Figure 19: Effect of cement alkalinity on pore water pH (theoretical pore water [OH-] lines are also shown).
As noted above, the high pH of concrete pore water is a result of the release to
alkali ions during cement hydration. Stochiometrically, the reaction of each mole of
cement alkali oxide produces two moles of hydroxide. Therefore, the theoretical
[OH-] in the concrete pore water is,
?MW50rEqAC
][OH-
⋅⋅⋅⋅
= , (16
where
C is the cement content of a concrete mix (kg/m3),
r is the degree of cement of hydration (fraction), and
MW is the molecular weight of Na2O (gms/mol).
Theoretical [OH-] values for concrete with C= 400 Kg/m3, r=1.00, and ρ=0.13, 0.15,
or 0.17 are plotted as lines in Figure 19. These lie above the measured values,
suggesting that full cement hydration had not been achieved with even the powdered
concrete samples exposed to distilled water for three days. It cannot be discounted,
however, that the OH- exhibited a binding effect similar to that of chlorides in
concrete. The data indicate that [OH-] for the HA mix concrete was approximately
0.0
0.20
0.40
0.60
0.80
1.00
0.0 0.20 0.40 0.60 0.80 1.00
Cement Equivalent Alkalinity, percent
Pore
Wate
r [O
H- ]
, M
W/C=0.37 W/C=0.41 W/C=0.50 Theoretical (ρ=0.13) Theoretical (ρ=0.15) Theoretical (ρ=0.17)
30
twice that of the NA, the latter being typical of Florida practice. Thus, the ratio of
[Cl-]-to-[OH-] for concretes of each of these two mixes that were similar except for
cement alkalinity and which underwent identical exposure should, for the HA mix
concrete, be approximately one-half that of the NA. If this ratio alone governs the
onset of passive film breakdown and active corrosion, then Cl- tolerance for the HA
concrete should be twice that of the NA. However, the data in Figure 17, where η
was greater for the LA concrete than for the NA, indicates that one or more other
factors was also involved as noted above.
Figure 20 compares the present [OH-] versus EqA results with ones available
from the literature. This indicates relatively good agreement between the two with
the present data occupying the lower bound of the results by others.
Figure 20: Relationship between pore water [OH-] and cement equivalent alkalinity as determined in the present and by past research.
Figure 21 plots the results from Figure 20 on the basis of pH and shows that this
parameter ranged from 13.05 to 13.20 for the LA mix, 13.19 to 13.36 for the NA,
and 13.53 to 13.65 for the HA.
Pore
Wate
r [O
H-]
, M
1.20
1.20
Cement Equivalent Alkaline, percent
0.0
0.20
0.40
0.60
0.80
1.00
1.20
0.0 0.20 0.40 0.60 0.80 1.00
Diamond (28,29) Arya (30)
Page (14) Constantiner (31) Tritthart (27)
Kawamura (32) Kayyali (33) Larbi (34)
Duchesne (35)
This Work
31
Figure 21: Relationship between pore water pH and cement equivalent alkalinity as determined in the present and by past research.
Chloride Analysis Results
Diffusion Coefficients
Figure 22 illustrates typical acid soluble [Cl-] analysis data as a function of
distance beneath the ponded concrete surface, as determined from a core acquired
from four of the blank (non-reinforced) specimens after 1,616 days of cyclic ponding
exposure. These results provide some insight into specimen-to-specimen scatter
(HA specimens), and indicate less Cl- penetration for this mix compared to the lower
equivalent alkalinity ones. Likewise, Figure 23 presents a plot of Deff versus EqA with
the data partitioned according to w/c. Cores from which the samples were acquired
for the Deff determinations were taken at either 1,210 or 1,616 days exposure except
for the single 0.50 datum for which this time was 133 days. The data indicate that
Deff for EqA = 0.97 was lower than for 0.52 by 44 percent. Figure 24 compares the
average Deff results from the present study with ones available from the literature
(30) and indicates that the present values are relatively high in comparison.
13.00
13.10
13.20
13.30
13.40
13.50
13.60
13.70
13.80
13.90
0.00 0.25 0.50 0.75 1.00 1.25 Cement Equivalent Alkaline, percent
pH
Literature This Work
32
0
4
8
12
16
20
0 25 50 75
Depth into Concrete, mm
Chlo
ride
Conce
ntr
atio
n,
kg/m
3
HA0.37-#6
HA0.37-#7
LA0.37-#48
NA0.37-#28
Figure 22: Example acid soluble [Cl-] data as a function of distance beneath the ponded concrete surface.
1.00E-13
1.00E-12
1.00E-11
1.00E-10
0.00 0.20 0.40 0.60 0.80 1.00 1.20
Equivalent Cement Alkalinity, percent
Eff
ective
Diffu
sion C
oef
., m
2/s
w/c=0.41w/c=0.37w/c=0.50
Figure 23: Effective diffusion coefficient data as a function of equivalent
cement alkalinity.
33
1E-13
1E-12
1E-11
1E-10
0.20 0.30 0.40 0.50 0.60 0.70 0.80 0.90
Water-Cement Ratio
Effe
ctiv
e D
iffu
sion
Coe
ffic
ient
, m
2/s
LiteratureThis Work
Figure 24: Comparison of Deff values from the present specimens with those
reported from the literature.
Figure 25 shows the Cl- profile that was determined for a core taken from a
previously tested LA0.50 specimen that was inverted approximately 50 months after
casting and re-ponded on what was initially the bottom surface for a single wet-dry
cycle. Also shown are profiles for LA0.41 and LA0.37 specimens (no LA0.50
specimen was available) after 80 wet-dry ponding cycles. Irrespective of the fact
that specimens of different w/c are involved, the profile for the one-cycle specimen
indicates that considerable Cl- uptake occurred early in the exposure, presumably as
a consequence of sorption. Thus, sorption could have been responsible for or
contributed to the relatively high Deff values that were calculated for the present
specimens (see Figure 24). The fact that sorption affected the [Cl-] profiles brings
into question the appropriateness of data analysis based upon Fick’s second law.
Chloride Threshold Concentrations
Table 6 lists [Cl-] values that were measured for powdered concrete samples
acquired from the top region of the upper rebar trace of specimens that had become
active, both at the active corrosion site and elsewhere where the steel remained
passive. Also listed are T i and the exposure time at which the specimens were
Mean+2σ
Mean-2σ
Mean
34
0
5
10
15
20
25
0 25 50 75
Distance Beneath Surface, mm
Chlo
ride
Conce
nta
tion,
kg/m
^3 LA0.50: 1-Cycle
LA0.37: 80-Cycles
LA0.41: 80-Cycles
Figure 25: Chloride profile after one wet-dry cycle compared to after 80 cycles.
autopsied. From these times and the Deff values (Table 7), [Cl-] at time T i was back-
calculated; and this value was taken as c th. Figure 26 shows a plot of cth versus Ti
and indicates that in most cases [Cl-] at the active corrosion site exceeded that
where the steel remained passive. However, in most of the cases where specimen
autopsy was performed at the time when potential first became negative to -0.28
VSCE, [Cl-] at the active and passive sites was approximately the same. This
suggests that the elevated [Cl-] at the corrosion site resulted from electromigration
subsequent to corrosion initiation. If this was the case, then it is the remote site [Cl-
] that should be taken as c th. No explanation is available for the data from Specimen
11-HA0.41, which is an exception to this trend. The general trend in Table 6 and
Figure 26 is one where c th increased with time, albeit at an ever decreasing rate,
irrespective of mix design. Thus, while specimens of the normal and low cement
alkalinity and high w/c exhibited the shortest Ti and vise versa, all data conformed to
a common trend. This is consistent with, first, chlorides having continued to
progressively accumulate with increasing time and, second, Ti being a distributed
parameter, as discussed above.
As noted above, c th values in the range 0.60-0.75 kg/m3 (1.0-1.3 pcy) have
35
Table 6: Exposure time data for specimens that have become active.
Specimen Number
[Cl-], kg/m3 (Active)
[Cl-], kg/m3 (Passive)
Time-to-Corrosion, days
Total Exposure Time, days
15 - HA 0.50 11.98 8.92 136 211
16 - HA 0.50 15.39 10.59 297 324
17 - HA 0.50 18.28 11.32 391 433
18 - HA 0.50 12.54 10.92 255 319
10 - HA0.41 23.27 23.53 1,192 1,459
11 -HA0.41 33.50 16.82 1,401 1,455
24 - NA0.37 10.93 10.76 1,096 1,124
25 - NA0.37 27.64 - 1,192 1,229
29 - NA 0.41 15.63 7.92 136 211
30 - NA 0.41 15.18 12.86 206 256
31 - NA 0.41 17.39 14.41 339 381
32 - NA0.41 18.76 - 1,120 1,186
36 - NA 0.50 19.67 10.93 93 183
37 - NA 0.50 19.56 6.90 93 213
38 - NA 0.50 16.62 9.13 136 256
39 - NA 0.50 10.22 7.29 178 211
45 - LA 0.37 16.90 11.22 149 211
50 - LA 0.41 10.44 10.95 297 324
52 - LA 0.41 21.46 16.49 567 576
58 - LA 0.50 11.87 5.42 234 264
59 - LA 0.50 15.61 4.31 136 211
60 - LA 0.50 20.33 7.37 136 256
64 - HAG 0.50 8.70 7.59 149 190
65 - HAG 0.50 7.84 6.87 178 211
66 - HAG 0.50 11.64 7.84 107 217
67 - HAG 0.50 13.99 9.18 163 211
historically been reported (1,2) from North America bridge deterioration deck
studies. Consistent with the projection above that c th is a function of multiple
factors, as well as being statistically distributed, Glass and Buenfeld (9) reported
values in the range 0.17-2.5 (total Cl- on a cement w/o basis) based upon literature
data. This translates to 0.66-9.71 kg/m3 (concrete weight basis) for the cement
content of the present specimens (Table 2). This upper limit from the Glass and
Buenfeld summary approximates the mid-value determined from the present
exposures. Possible explanations for the present relatively high c th values include
the following:
36
Figure 26: Plot of Time-to-Corrosion as a function of [Cl-] for specimens that
have become active.
1. Specimens were exposed in laboratory air at a constant relative humidity
(RH) of about 70 percent.
2. The absence of RH and temperature variations.
3. Chloride binding effects.
4. Surface condition of the reinforcement (wire brushed as opposed to as-
received).
5. The passive film upon the steel may become more stable with time.
6. Concrete aging with time.
7. Specimens were relatively small such that the nature of the steel-concrete
interfacial may not reflect statistically the variability that can arise in large
structures.
With regard to Items 1 and 2, it can be reasoned that temperature/relative humidity
variations, as occur with outdoor conditions, promote fluxes of moisture and oxygen
Time-to-Corrosion, days
Chlo
ride
Conce
ntr
ation,
kg/m
3
0
10
20
30
40
0 200 400 600 800 1000 1200 1400 1600
Corrosion Spot Remote Spot
37
in concrete that do not take place under controlled laboratory conditions and that
these, in turn, facilitate the onset of corrosion. Experiments are ongoing to
investigate this further.
Chloride binding (Item 3) reduces the concentration of this species that is “free”
and, hence, available to participate in steel depassivation and corrosion. While it is
reasonable to assume that some fraction of the total chlorides that migrated into the
concrete specimens became bound, it is unlikely that the concentration of these was
sufficient to account for the difference between the measured c th for the present
experiments (4.3-23.5 kg/m3 (7.3-39.5 pcy), see Table 8) compared to historically
reported values for North America bridge concretes (1,2), given that, first, this
difference is by a factor of 7-40 and, second, C3A concentration for all three cements
was in the normal range (see Table 1).
Li and Sagüés (8) found, based upon exposure of carbon steel in simulated pore
water solutions (pH ~ 12.6-13.6), that cth for as-received and pre-rusted specimens
was approximately twice that of sandblasted ones. If performance of specimens with
a wire brushed surface finish, as was the case for the present experiments, is
comparable to sandblasted and if results from aqueous exposures can be
quantitatively compared to those for steel in concrete, then this factor (surface
preparation) probably contributed to about a factor of two of the difference between
the present and previously reported (1,2,9) c th values.
Items 5 and 6 were addressed by a series of experiments where the G109
specimens that had become active were retested. This involved reconfiguring these
by inverting, grouting together the split specimen halves, reponding, and monitoring
potential of what was, upon initial exposure, the two bottom bars for onset of active
corrosion. The regrouting was intended to provide mechanical stability to the
specimens, and no electrical connection was made between the top and bottom bars.
Figure 28 shows a photograph of two of these specimens. The specimens were
originally fabricated over a one month period in late-1997, and the original ponding
commenced in July, 1998 (approximate concrete age eight months). The reponding,
on the other hand, began in June, 2001 (approximate concrete age 43 months). It
was reasoned that if either enhanced passivity with time or concrete aging caused or
contributed to the observed c th versus time trend, then the results upon reponding
38
Figure 27: Photograph of two repaired and inverted specimens under test.
should exhibit greater T i and cth than occurred upon the initial exposure.
Figure 28 presents the Weibull formatted T i results upon reponding for the
HA0.50 specimens in comparison to data for the original exposures. Here, the β
value for the two data sets is approximately the same; however, η was less for the
reponded tests than for the originals by a factor of approximately three. Results for
the NA0.50 and LA0.50 specimens were similar, although in these cases the
difference in η was by a factor of 1.8 and 2.5, respectively. On the other hand, η for
the reponded HAG specimens was slightly greater than for the original tests. The
relatively small number of NA0.41 and LA0.41 specimens in either category (original
or retest) that have become active precluded realistic comparisons here, although
the limited results are consistent with η being less for the reponded exposures. On
this basis, Items 5 and 6 were both discounted as being possible explanations for the
relatively high c th values and the variation with time.
With regard to Item 7 (size effect), it is generally recognized in failure analysis
that the probability of failure increases with increasing member size. When
mechanical fracture is involved, a contributing factor is that strength often decreases
with increasing section thickness because of fabrication effects. Also at issue is the
fact that the probability of a critical size defect being present increases with
increasing size. On this basis, a relatively small G109 specimen is less likely to
Repaired sectioning of
concrete along the line of what was the top bar.
Two monitored bars of each specimen.
39
Figure 28: Weibull distributions of T i for the original and retested exposures of HA0.50 specimens.
exhibit a corrosion initiating feature or defect of a specific significance than is a large
slab or structure. Exactly what type of defect(s) or microstructural feature(s) could
serve such a role in the case of reinforcing steel in concrete is uncertain; but
possibilities include 1) the heterogeneous nature of concrete such that some specific
condition is locally established at the steel-concrete interface, 2) a preexisting
geometrical feature upon the steel surface (locally severe surface roughness or a
micro-crevice, for example), and 3) a pit-promoting microstructural feature at the
steel surface (inclusions, for example). No effort was made to investigate 2) and 3);
however, as a matter of hindsight, it was possible to discern information that may
pertain to 1).
Previous research has reported that corrosion of reinforcing steel preferentially
initiates at concrete voids that impinge upon the steel (37). Visual observation of
the initial specimens that became active in the present project revealed one or more
air voids at the site of active corrosion. Figure 29 shows an example of this for
Specimen 59 (LA0.50). Any such voids at the corrosion site of specimens that
became active later were not detected because of the build-up of corrosion products
here. Further into the program, however, the question of corrosion initiating at voids
40
Figure 29: Photograph of rebar trace at the active corrosion site for Specimen 59 (LA0.50) after autopsy showing concrete voids at or near the corrosion epicenter.
in the concrete was revisited. In this regard, Figure 30 shows a general view of
Specimen 10 (NA0.41) subsequent to sectioning; and Figure 31 presents a close-up
photograph of the corrosion site that was taken shortly after sectioning.
Deliquescent droplets that are apparently hydrolyzed corrosion products of low pH
are apparent here. Figure 32 shows a close-up view of the rebar trace in the
concrete at the corrosion site opposite the view in Figure 31. This reveals a void in
the concrete at the corrosion site (circled area) corresponding to a location of
deliquescence in Figure 31.
Specimen 10 was sectioned further, and the rebar trace at the corrosion site was
viewed using a scanning electron microscope and chemically analyzed via Energy
Dispersive X-ray Diffraction (EDX). In this regard, Figure 33 reproduces Figure 32
with various void areas that were analyzed identified. Correspondingly, Table 7 lists
the chlorine, presumable as Cl-, analysis results for these and shows that this species
was present at almost 22 w/o at the void where corrosion had initiated. This high
value presumably resulted from electromigration, as discussed above. Such a
concentration is consistent with the deliquescent nature of the corrosion products.
The average [Cl-] for the other voids was approximately one w/o. Even this is a
Void (2)
Rebar Trace
41
Figure 30: Photograph of Specimen 10 (HA0.41) after sectioning showing active localized corrosion on the upper portion of the top bar.
Figure 31: Photograph of the locally corroded region on the exposed top half
of the upper rebar of Specimen 10 (HA0.41) after sectioning. relatively high value and may reflect a tendency for chlorides to concentrate at voids
or the relatively poor accuracy of EDX at low concentrations. It was not possible to
analyze other specimens in a similar fashion because powder drilling had already
been acquired from the corrosion site on these.
42
Figure 32 Close-up view of the rebar trace in the concrete at the corrosion site of Spec imen 10 (HA0.41).
Figure 33: Photograph of the rebar trace from Specimen 10 (HA0.41) showing concrete voids that were compositionally analyzed.
Table 7: Chloride analysis results for the voids identified in Figure 34.
Location Void-C Void-R1 Void-R2 Void-R3 Void-R4 Void-R5
21.88 1.43 0.66 - 0.93 1.87
Void-
Void-R1
Void-R2
Void-R3
Void-R4
Void-R5
43
Based upon the above observations, it is concluded that concrete voids may
have facilitated passive film breakdown and onset of localized corrosion for the
present specimens. If this was the case, then the mechanism probably involved a
concentration cell with contiguous concrete coated and bare steel serving as cathode
and anode, respectively. In this regard, it has previously been reported (38) that
such a cell has a relatively high driving potential and is likely to be compounded by
an unfavorable surface area ratio (small anode–large cathode). It can be reasoned
that the specimen of a given mix design with the largest void intersecting the top
portion of the upper bar activated first and vise versa. On this basis, the variability
in c th (Table 8 and Figure 26) resulted in response to probabilistic aspects of void
size, density, and distribution.
If voids were a factor in corrosion initiation for the present specimens, then the
possibility exists that durability of concrete structures could be enhanced by any
treatment that reduces their size and density. This might be affected by 1)
innovative placement procedures, 2) admixtures that are specially formulated for
void size/density reduction, and 3) employment of self-consolidating concrete.
Additional specimens must be analyzed microscopically and additional exposures
performed before definitive conclusions can be reached regarding this possibility.
ACKNOWLEDGEMENT
The authors are indebted to Mr. Rodney Powers of the FDOT Corrosion
Laboratory for fabricating the specimens and to Mr. George Jones for assistance
with various aspects of the experiments.
CONCLUSIONS
The following conclusions were reached based upon exposure of a series of G109
concrete specimens to cyclic wet-dry ponding with a 15 w/o NaCl solution.
1. Time-to-corrosion for “identical” specimens of each mix design varied, often
over a relatively wide range. This resulted because this parameter, time-to-
corrosion, is statistically distributed and is effected by a number of factors,
each of which is also distributed.
44
2. Of the Type II cements that were employed in specimen fabrication, the
exposure time for passive film breakdown and onset of active corrosion, T i,
was least for ones that that were of “normal” alkalinity (equivalent alkalinity
0.52 (designation NA)), intermediate for a “low” alkalinity cement (equivalent
alkalinity 0.36 (designation LA)), and highest for “high” alkalinity (equivalent
alkalinity 0.97 (designation HA)). Because the equivalent alkalinity of
cements typically utilized in Florida bridge construction is comparable to that
of the NA type and because alkali silica reaction is not normally a problem in
the State, improved corrosion resistance can be realized using cements of the
HA type. The increase in time-to-corrosion for specimens with the HA cement
compared to the NA ones was by approximately a factor of three for water-
cement ratio 0.37 and by a factor of ten for water-cement ratio 0.41. Time-
to-corrosion did not increase monotonically with cement alkalinity in that Ti
was greater for specimens with the LA cement than with the NA. This is
thought to have resulted from relatively high Cl- binding in the lower pore
water pH range.
3. Pore water pH was in the range 13.05-13.20 for the LA mix concrete, 13.19-
13.36 for the NA, and 13.53-13.65 for the HA. This resulted from OH-
concentration for the HA cement concrete being about a factor of two greater
than for the NA. The pH-cement alkalinity trend for the present concrete
mixes was slightly above the lower bound of data in the literature.
4. The threshold chloride concentration to cause passive film breakdown and
active corrosion, c th, increased with increasing T i. Consequently, the HA
specimens exhibited the highest c th although all data conformed to a common
trend. Local chloride concentrations in the concrete measured from powder
drillings taken along the top side of the upper rebar trace varied from 7.84 to
27.64 kg/m3 (concrete w/o basis) at the site of active corrosion and from
4.31 to 23.52 kg/m3 at locations where the steel was passive. This difference
presumably resulted from electromigration of Cl- to active sites subsequent to
corrosion initiation. The latter range (4.31 to 23.52 kg/m3) is considered to
represent cth, since Cl- electromigration and rapid accumulation of this species
subsequent to corrosion initiation should not have occurred here. This range
45
(4.31 to 23.52 kg/m3) exceeds what is typically assumed to initiate
reinforcement corrosion in North American bridge deck concretes (0.60-0.75
kg/m3 (1.0-1.3 pcy)) by a factor of 6-39. On the other hand, the general
literature reports the range for c th as 0.66-9.71 kg/m3 which, while
overlapping, is still less than determined for the present specimens.
5. The effective diffusion coefficient, Deff, increased with increasing water-
cement ratio and was about 40 percent lower for concrete prepared with the
HA cement compared to the NA and LA ones. The present Deff values are at
the upper bound or exceed those reported in the literature. Results from
short-term exposures indicated that sorption was a significant contributor to
Cl- uptake by the present specimens, and this apparently inflated the Deff
determinations. This role of sorption also brings into question the
appropriateness of Fick’s second law for analyzing the present Cl- data.
6. In situations where observations were made, corrosion was found to have
initiated at a void in the concrete that impinged upon the reinforcement. A
concentration cell whereby the contiguous concrete coated and bare steel
served as cathode and anode, respectively, is considered to have been
responsible. The size and density distribution of these voids may have
contributed to or been responsible for the finding that T i was a distributed
parameter.
RECOMMENDATIONS
Based upon the results of this research it is recommended that further
research be performed to determine the following:
1. Feasibility of the private sector producing high alkalinity cements suitable
for Florida bridge construction. Such a study should consider a)
identifications of problems that industry might encounter in accomplishing
this, b) different options for attaining high alkalinity, and c) definition of
an appropriate alkalinity range for corrosion resistance enhancement.
2. Confirmation of the role of air voids in corrosion initiation. Activities in
this category, subsequent to confirmation, should address a) possible
46
admixtures for void elimination and b) the utility of self-compacting
concretes for void reduction.
BIBLIOGRAPHY
1. Spellman, D.S. and Stratfull, R.F., “Concrete Variables and Corrosion Testing,” Highway Research Record, No. 423, 1963, p. 27.
2. Clear, K.C., “Time-to-Corrosion of Reinforcing Steel in Concrete Slabs,” Report
No. FHWA-RD-76-70, Federal Highway Administration, Washington, D.C., 1976. 3. Hausmann, D.A., Materials Protection, Vol. 6(10), 1967, p. 23. 4. Gouda, V.K., British Corrosion Journal, Vol. 5, 1970, p. 198. 5. Hausmann, D.A., Materials Performance, Vol. 37(10), 1998, p. 64. 6. Breit, W., Mater. Corrosion, Vol. 49, 1998, p. 539. 7. Charvin, S., Hartt, W.H., and Lee, S. K., “Influence of Permeability Reducing and
Corrosion Inhibiting Admixtures in Concrete upon Initiation of Salt Induced Embedded Steel Corrosion,” paper no. 00802 presented at CORROSION/00, March 26-31, 2000, Orlando.
8. Li, L. and Sagüés, A.A., Corrosion, Vol. 57, 2001, p. 19. 9. Glass, G.K. and Buenfeld, N.R., “Chloride Threshold Levels for Corrosion Induced
Deterioration of Steel in Concrete,” paper no. 3 presented at RILEM International Workshop on Chloride Penetration into Concrete, Oct. 15-18, 1995, Saint Rémy-les-Chevreuse.
10. Broomfield, J.P., Corrosion of Steel in Concrete, E&FN Spon, London, 1997, pp.
22-25. 11. Tutti, K., Corrosion of Steel in Concrete, Report No. Fo 4, Swedish Cement and
Concrete Research Institute, Stockholm, 1982. 12. Longuet, P., Burglen, L., and Zelwer, A., Rev. Matér., Vol. 35, 1973, p. 676. 13. Barneyback, R.S. and Diamond, S., Cem. Concr. Res., Vol. 11, 1982, p. 279. 14. Page, C.L. and Vennesland, O., Mater. Struct., Vol. 19, 1983, p. 16. 15. Sagüés, A.A., Moreno, E.I., and Andrade, C., Cem. Concr. Res., Vol. 27, 1997, p.
1747. 16. Li, L., Sagüés, A.A., and Poor, N, Cem. Concr. Res., Vol. 29, 1999, p. 315. 17. Hartt, W.H., Charvin, S., and Lee, S.K., “Influence of Permaeability Reducing and
Corrosion Inhibiting Admixtures in Concrete upon Initiation of Salt Induced
47
Embedded Metal Corrosion,” final report submitted to FDOT by FAU on Project WPI 0510716, June 18, 1999.
18. Bamforth, P.B., “Definition of Exposure Classes and Concrete Mix Requirements
for Chloride Contaminated Environments,” Corrosion of Reinforcement in Concrete, Eds: Page, C.L., Treadaway, K.W.J., and Bamforth, P.B., Soc. Chem. Ind., London, 1996, p. 176.
19. Bamforth, P.B. and Price, W.F., “Factors Influencing Chloride Ingress into Marine
Structures,” Concrete 2000, Eds: Dhir, R.K. and Jones, M.R., E&FN Spon, London, 1993, p. 1105.
20. ASTM Standard Test Method G 109-99, “Determining the Effect of Chemical
Admixtures on the Corrosion of Embedded Steel Reinforcement in Concrete Exposed to Chloride Environments,” Annual Book of ASTM Standards, Vol. 03.02, American Society for Testing and Materials, 100 Barr Harbor Drive, West Conshohocken, PA.
21. “Florida Method of Test for Determining Low-Levels of Chloride in Concrete and Raw Materials,” Designation FM 5-516, Florida Department of Transportation, Tallahassee, FL, Sept., 1994.
22. Li, L., Nam, J., and Hartt, W.H., “Ex-Situ Leaching Measurement of Concrete
Alkalinity,” paper no. 03299 to be presented at CORROSION/03. 23. ASTM Standard Test Method C 642-90, “Test Method for Density, Absorption, and
Voids in Hardened Concrete,” Annual Book of ASTM Standards, Vol. 04.02, American Society for Testing and Materials, 100 Barr Harbor Drive, West Conshohocken, PA.
24. ASTM Standard Test Method G 876-99, “Test Method for Half-Cell Potentials of
Uncoated Reinforcing Steel in Concrete,” Annual Book of ASTM Standards, Vol. 03.02, American Society for Testing and Materials, 100 Barr Harbor Drive, West Conshohocken, PA.
25. Hartt, W.H., Nam, J., and Li, L. “A Unified Approach to Concrete Mix Design
Optimization for Durability Enhancement and Life-Cycle Cost Optimization,” Draft Final Report submitted to FDOT Research Center by Florida Atlantic University, June 13, 2002.
26. Sandberg, P. and Larsson, J., “Chloride Binding in Cement Pastes in Equilibrium
with Synthetic Pore Solutions,” in, Chloride Penetration in Concrete Structures, Ed. L.-O Nilsson, Nordic Miniseminar, Göteborg, 1993, p. 98.
27. Tritthart, J., Cem. Concr. Res., Vol. 19, 1989, p. 683.
28. Diamond, S. Cem. Concr. Res., Vol. 11, 1981, p. 383. 29. Diamond, S., “Effects of Microsilica (Silica Fume) on Pore Solution Chemistry of
Cement Pastes,” Communications of Am. Ceramics Soc., 1983, p. C82.
30. Arya, C. and Xu, Y. Cem. Concr. Res., Vol. 25, 1995, p. 893.
48
31. Constantiner, D. and Diamond, S., “Pore Solution Analysis: Are There Pressure Effects,” in Mechanisms of Chemical Degradation of Cement-based Systems, Eds: K.L. Scrivener and J.F. Young, E&FN SPON, London, 1997, p. 22.
32. Kawamura, M., Kayyali, O.A., and Haque, M.N., Cem. Concr. Res., Vol. 18, 1988, p. 763.
33. Kayyali, O.A. and Haque, M.N., J. of Materials in Civil Engr., Vol. 2, 1990, p. 24.
34. Larbi, J.A., Fraay, A.L., and Bijen, J.M., Cem. Concr. Res., Vol. 20, 1990, p. 506.
35. Duchesne, J. and Berube, M.A., Cem. Concr. Res., Vol. 24, 1994, p. 456.
36. Bamforth, P., Corrosion Prevention and Control, Aug., 1996, p. 91.
37. Monfore, G.E. and Verbeck, G.J., ACI Journal Proceedings, Vol. 32, 1960, p. 491.
38. Miller, R.L., Hartt, W.H., and Brown, R.P., Materials Performance, Vol. 15(5), 1976, p. 20.